[BioC] two channel data normalization limma

[Gordon K Smyth]

Dear Maite,

You write "I read that for time series experiments you should do single 
channel analysis".  Where have you read this?  I see no reason why you 
need to do two different types of analysis of the same data.

The limma User's Guide gives a lot of guidance regarding how to normalize 
data, whether you use a log-ratio or a separate channel approach.

Best wishes
Gordon

PS. Please do not post the same question to Bioconductor multiple times. 
I got your question 3 times.

---------------------------------------------
Professor Gordon K Smyth,
Bioinformatics Division,
Walter and Eliza Hall Institute of Medical Research,
1G Royal Parade, Parkville, Vic 3052, Australia.
http://www.statsci.org/smyth

On Thu, 12 Jul 2012, Maite Iriondo [guest] wrote:

>
> Dear Bioconductor Community,
>
> I am Maite Iriondo and I am currently finishing a MSc programme, doing 
> my thesis in microarray data analysis even though I lack of 
> bioinformatic background as I come from Food Science. I am trying to 
> understand the different options for normalizing two channel microarray 
> data, and I have very basic questions that I would appreciate if they 
> could be answered.
>
> 1- I have read in the literature that two channel data can be analized 
> by single channel and two channel normalization approaches (using log 
> intensities and log ratios respectively) (Yang and Thorne, 2003). 
> However, I do not understand how these approaches are applied in the 
> limma package (Bioconductor). For two channel normalization you use a 
> within array and between array normalization and for single channel 
> analysis you use just the options of the between array normalization 
> function?
>
> 2- My data was partially analised with one scanner and then changed to 
> another, which gives huge differences on the widths of the boxplots. I 
> read in Smyth and Speed (2003) that scale normalization between arrays 
> is recommended. Which function in limma does this calculation? is it the 
> normalizeBetweenArrays (method= scale)?
>
> 3- My data consists of a series of microarrays from P.putida growth 
> experiments at different temperatures, taking samples at different time 
> points for each temperature. Also, a second experiment was carried doing 
> microarrays using adapted and non adapted cells to low temperatures of 
> P. putida at different temperatures and again different time points for 
> each temperature. As a summary:
>
> A) Microarrays at 30, 10 and 5ºC with 5 time point samples (eg. At 
> 30ºC I have data from 5 microarrays that correspond to 5 time points in 
> which the cells were taken) B) Microarrays at 5 and 10ºC using adapted 
> and non adapted cells with 5 time point samples
>
> In the beginning I normalized all the data together using different 
> options of normalizeWithinArrays() and normalizeBetweenArrays(), but the 
> resulting MAplots and boxplots dont look too good. I was wondering if I 
> should separate experiments A and B. Also, I read that for time series 
> experiments you should do single channel analysis. But for comparing 
> gene expression between different temperatures I should use two channel 
> analysis. This means that I should analyse each temperature separately 
> using a single channel approach and then use a two channel approach for 
> all the data to compare gene expression between temperatures?
>
> Thank you for your time and effort in advanced. I know that these 
> questions are very basic but I am trying to put my concepts in order.
>
> Maite

______________________________________________________________________
The information in this email is confidential and intend...{{dropped:5}}